fetch-embed 0.1.6

No project description provided

fetch-embed

[]

fetch multilingual embed from embed.ttw.workers.dev

Install it

pip install -U fetch-embed

Use it

Make use of the helper function embed_text

endpoints for two models (dim-52 and dim-768)

via cloudflare: https://embed.ttw.workers.dev/embed/ and https://embed.ttw.workers.dev/embed_la/

In case you cannot access embed.ttw.workers.dev, you may use ttw.hopto.org (hosted by noip.com) instead.

Swagger UI: Self-docs for these endpoints

https://embed.ttw.workers.dev/docs

https://ttw.hopto.org/docs

Model 1: multilingual, dim-512

The default endpoint is https://embed.ttw.workers.dev/embed/

from fetch_embed.embed_text import embed_text

res = embed_text(["test a", "测试"])
print(res.shape)
# (2, 512)

Model 2: language agnostic, dim-768

endpoint: https://embed.ttw.workers.dev/embed_la/

from fetch_embed.embed_text import embed_text

endpoint = "https://embed.ttw.workers.dev/embed_la/"

res = embed_text(["test a", "测试"], endpoint=endpoint)
print(res.shape)
# (2, 768)

Consult the embed_text.__doc__ (e.g. print(embed_text.__doc__)) or its source code for more details.

Access the API directly

from fetch_embed import fetch_embed

res = fetch_embed("test me")
print(res.shape)
# (1, 512)

print(fetch_embed(["test me", "测试123"]).shape
# (2, 512)

# to turn off live progress bar
res = fetch_embed("test me", livepbar=False)

# brief docs
help(fetch_embed)
# fetch_embed(texts:Union[str, List[str]], endpoint:str='http://ttw.hopto.org/embed/', livepbar:bool=True) -> numpy.ndarray
    Fetch embed from endpoint.

Plug in endpoint = "https://embed.ttw.workers.dev/embed_la/" for Model 2, e.g.,

import numpy as np
from fetch_embed import fetch_embed

endpoint = "https://embed.ttw.workers.dev/embed_la/"
res = fetch_embed("test me", endpoint=endpoint)
print(np.array(res).shape)
# (1, 768)